Decline of Large-Diameter Trees in a Bamboo-Dominated Forest Following Anthropogenic Disturbances in Southwestern Amazonia

Annals of Forest Science (2019) 76:110 https://doi.org/10.1007/s13595-019-0901-4

RESEARCH PAPER

Decline of large-diameter trees in a bamboo-dominated forest following anthropogenic disturbances in southwestern Amazonia

Leonardo G. Ziccardi1,2 & Paulo Maurício Lima de Alencastro Graça1 & Evandro O. Figueiredo3 & Philip M. Fearnside1

Received: 3 May 2019 /Accepted: 17 November 2019 # INRA and Springer-Verlag France SAS, part of Springer Nature 2019

Abstract & Key message Reduction in the aboveground biomass of larger trees is the main consequence of disturbances in open forests dominated by bamboo. Because these trees are of central importance both for ecosystem function and for the economic value of the forest for management, the impact on these trees due to the increase of bamboo abundance following anthropogenic disturbances is both an environmental and a commercial concern. & Context Bamboo-dominated forests in southwestern Amazonia are increasingly exposed to the combined impacts of fire and selective logging. Although the climbing bamboos (Guadua spp.) are considered native species of these forests, disturbances contribute to a discontinuous canopy, which is an ideal scenario for an increase in abundance of these opportunistic plants. The regeneration of tree biomass is limited in these cases, decreasing the carbon stock of the forest. & Aims This study compares changes in bamboo abundance and forest structure in the face of forest fire and post-burn logging in the municipality (county) of Rio Branco, Acre state, Brazil. & Methods The study was conducted on the Transacreana Highway (AC 090) in the eastern portion of Acre state, Brazil. We compared changes in bamboo abundance and aboveground biomass (AGB) of an area of forest with no known recent disturbances to areas in the same forest that had been disturbed by fire and post-burn logging, which are frequent sources of disturbance in this region. The live and dead AGB values were estimated by field inventory in 2016, which was 11 years after a fire and 9 years after selective logging. The AGB values for trees and palms were estimated by allometric equations in three 2-ha areas. Bamboo abundance was expressed both by bamboo biomass (Mg ha−1) and culm density (culms ha−1), sampled by direct measurements in three 200-m2 areas. & Results Bamboo abundance was over 20% higher in forest that had been burned but not logged, as compared to undisturbed forest. Compared to undisturbed plots, aboveground forest biomass (live + dead) was 34% lower in plots exposed to fire but not logging, and 36% lower if exposed to both. Live trees and palms represented 77% of the total forest biomass (live + dead), and almost a half of this contribution (41%) was in individuals in the largest diameter class (diameter at breast height > 50 cm). & Conclusion Increase in the level of impact led to a reduction in aboveground carbon stock of the forest. One of the main consequences of disturbances in open forests dominated by bamboo is reduction of live trees and palms, especially in the diameter at breast height (DBH) class over 50 cm.

Keywords Brazil . Acre . Bamboo forest . Guadua spp. . Forest fires . Logging

Handling Editor: Paulo Fernandes Contributions LGZ designed the study and conducted the fieldwork, data analysis and writing. PMLAG, EOF and PMF contributed to the project design, data interpretation and writing

* Leonardo G. Ziccardi 2 Department of Forestry, Michigan State University, East [email protected] Lansing, MI 48824, USA 3 Embrapa Acre, Rodovia BR-364, km 14, CEP 69900-056, Rio 1 Department of Environmental Dynamics, National Institute for Branco, Acre, Brazil Research in Amazonia (INPA), Av. André Araújo, 2936, CEP 69067-375, Manaus, Amazonas, Brazil 110 Page 2 of 13 Annals of Forest Science (2019) 76:110

1 Introduction tree species of commercial value becomes limited (D’Oliveira et al. 2013;Rockwelletal.2014). Forest fires are not common natural events in the Amazon Occurrence of forest fires is critical under these cir- rainforest. The only known natural cause of forest fires is light- cumstances, because an increase in tree mortality, in bam- ning strikes, and in humid tropical forests, burning is almost boo dominance, and in dead biomass (necromass) pro- always limited to a single tree or a very small patch of vegeta- vides fuel for future fire events (Smith and Nelson 2011; tion (Fearnside 1990). In contrast, forest fires from human ig- Veldman and Putz 2011). Intensification of climate nition are increasingly frequent sources of disturbance. change, with effects such as more-frequent extreme Fire occurrences are directly related to the vegetation type, droughts, and increases in other direct human impacts, climatic conditions and socioeconomic context of a given re- such as selective logging, is consistent with the observed gion. Even at different time scales, a strong relationship persists progressive increase in the number of forest fires and of between the fire regime and the territorial dynamics (Molina the gaps that are left as a legacy of these disturbances and Galiana-Martín 2016). Logging is an example of human killingtreesandpalms,enhancingbamboodensityin activity that determines forest-fire susceptibility. Logging roads these forests (Lewis et al. 2011;Ferreira2014). and the gaps formed by felling trees promote the opening of the The increase in culm density per unit area may not necessarily canopy, contributing to an increase in the probability of occur- reflect a direct increase in bamboo biomass. Culms tend to lose rence of forest fires (Nepstad et al. 1999). verticality in forests with reduced numbers of large trees, where An important reflection of anthropogenic disturbances can be bamboo becomes locally dominant and has diameter increments observed in the open forests of southwestern Amazonia. These of ramets restricted by limited access to light (Yavit 2017). This forests occupy an estimated area of at least 161,500 km2 and are scenario of bamboo dominance in a forest leads to alterations in characterized by the dominance of native bamboo species of the the floristic composition, reducing the number of species in a genus Guadua in the understory. Although the occurrence of single hectare by almost 40%, on average (Silveira 2005). Guadua spp.in the Amazon is naturally associated with luvisols An important first step to reduce degradation effects in (Kandisols) and eutrophic cambisols (Inceptisols), disturbance in these forests is to better understand and quantify the impacts the forest can promote clearings favorable to Guadua spp., cre- associated with the main sources of disturbance. This paper ating conditions for it to occur in areas where it would not oth- aims to help fill an important gap in knowledge of the effect of erwise exist (de Carvalho et al. 2013;Ferreira2014). Forests of fire and post-burn logging on the structure and composition of southwestern Amazonia subject to disturbances such as wind- forests dominated by Guadua spp. in the understory. Here, we throws, logging, and fire are characterized by a discontinuous investigate how these disturbances affect the dominance of canopy that provides ideal conditions for invasion of aggressive bamboo, aboveground forest biomass, and abundance of large pioneering plants, such as bamboo. trees in southwestern Amazon forests. Open bamboo-dominated forests have a reduced number of trees in the largest diameter classes when compared to forests where bamboo does not occur, which promotes a reduction in 2 Materials and methods total aboveground biomass, carbon stock, and, consequently, in the economic value of these forests (Oliveira 2000; Nelson 2.1 Study area et al. 2001). Avoiding degradation in these forests is essential for the continuity of important ecosystem services (ES), such The study area is located in the eastern portion of Acre State on as hydrological regulation, maintenance of biodiversity, and the Transacreana Highway (AC 090), 45 km from the city of Rio carbon stock (Daily 1997). Branco (9° 59′ 10″ S; 68° 11′ 33″ W). The municipality (county) Logging is not likely to decrease in the near future in of Rio Branco has an area of 8835.5 km2 and an estimated Brazil, given the new presidential administration’sdra- population of 383,443 (Brazil, IBGE, 2017). The region has an matic shift toward less environmental regulation and more annual average temperature of 25.6 °C and receives mean annual promotion of extractive activities (Ferrante and Fearnside precipitation of approximately 2200 mm (Sombroek 2001). The 2019).Logginghasbeenshowntoformgapsbyremoval rainy season usually begins in October and lasts until May. The oflargetreesfromtheforest.Thesegapsprovidefavor- rainiest quarter comprises the months of January, February, and able conditions for native climbing bamboo, often starting March, while the driest quarter corresponds to June, July, and a self-perpetuating disturbance cycle where these bam- August (Acre, Governo do Estado 2000). boos hang on the trees, causing physical damage to the Two areas were selected in the same forest remnant, both their crowns and promoting greater canopy opening, disturbed by the same forest fire in 2005. One of the areas was which is a favorable scenario for the development of more further disturbed by logging in 2007 (logging intensity ramets (Griscom and Ashton 2006). Thus, regeneration of 16.82 m3 ha−1). Based on the 0.54 g cm−3 average wood density value in this part of Amazonia (Nogueira et al. Annals of Forest Science (2019) 76:110 Page 3 of 13 110

Fig. 1 Location map of study areas (c) in the municipality of Rio Branco (b), Brazil (a)

2007), the loss of wood biomass after logging was 100-m intervals, summing 2 ha as the sample area for estimating 10.16 Mg ha−1. In this same remnant, approximately 4 km aboveground biomass in each treatment (Fig. 2). to the south of the first two areas, we selected an area used Bamboo biomass was obtained directly in 8 subplots of 5 × as reference (control) that was not affected by fires or by 5m(25m2), installed systematically, totaling a sample area of recent extractive processes (Fig. 1). 200 m2 for each treatment. According to local residents, most of the forests in the region were disturbed by forest fires in 2005. A major drought 2.3 Aboveground biomass occurred in 2005 in this part of Amazonia (da Silva et al. 2018; Vasconcelos et al. 2013), and the many cattle ranches along Aboveground biomass was estimated by summing live and dead the AC-90 Highway probably also contributed to the wide- biomass (necromass) classes. All values referring to estimates of spread fires. The increase of the density of bamboo in the live and dead biomass of trees and palms were obtained by allo- forest was subsequent to these fires. metric equations developed for the Amazon region. Bamboo biomass was obtained by direct weighing. Woody residues were es- 2.2 Sampling design timated by line-intercept transects (Van Wagner 1968)(Table1).

The study was carried out along three transects, each 1100 m in 2.3.1 Live trees and palms length (one for each treatment). Placement of transects respected a minimum distance of 100 m from the edge of the forest. Along For aboveground biomass estimation of live trees, we used the each transect, 8 plots of 50 × 50 m were systematically placed at equation developed for forests located in the “arc of deforesta-

Fig. 2 Schematic of the sampling design in each transect of 1100 m in length 110 Page 4 of 13 Annals of Forest Science (2019) 76:110

Table 1 Allometric equations and methods used to estimate Compartment Equation/method Author biomass in each compartment Live trees (DBH ≥ 10 cm) [ln (AGB) = −1.716 + 2.413 × ln Adapted from (Nogueira (DBH)] × c et al. 2008a) Live palms (DBH ≥ 10 cm) [ln (AGB) = −3.3488 + 2.748 × [ln Adapted from (Goodman (DBH)] × p et al. 2013) Standing dead trees (DBH ≥ 10 cm) AGB = G × H × β0 ×d×0.1 Brownetal.1995 Crownless trees and dead palms AGB = G × H × F × d Graça et al. 1999 (DBH ≥ 10 cm) Woody debris (DBH ≥ 10 cm) Line-intercept transects Van Wagner 1968 Bamboo (live and dead) Direct weighing –

AGB aboveground biomass (dry weight in kg); DBH diameter at breast height, measured 1.30 m above the ground or above any buttresses (cm); c species wood density correction coefficient for trees (wood density of the species divided by the mean wood density in the area, which for this part of Amazonia is 0.540 g cm−3 ) (Nogueira et al. 2008a); p density correction coefficient for palms. Mean wood density = 0.488 g cm−3 (Chave et al. 2006); G cross-sectional area (cm2 ); H total height (m); β0 regression coefficient (0.62); d density attributed to the decomposition phase (g cm−3 ); F form factor, or the ratio of the volume of a bole to the volume of a cylinder with the length of the bole and diameter of the DBH (0.78) tion” along the southern edge of the Amazon forest (Nogueira biomass of standing dead trees, crownless trees, dead palms, et al. 2008a), incorporating a density correction coefficient and woody debris. basedontheratiobetweenthespecies wood density (Zanne −3 et al. 2009) and the average value of 0.54 g cm for wood 2.3.3 Standing dead trees and palms density in this region (Nogueira et al. 2007). The same density correction coefficient was used to calculate the ratio between Biomass of standing dead trees was quantified using the the palms species wood density and the average value of −3 equation fit by Brown et al. (1995) at the Samuel ecolog- 0.488 g cm for the family Arecaceae (Chave et al. 2006). ical station (Rondônia) and adapted by Nascimento and For species that did not have densities in the database, we Laurance (2006) when considering wood density. We used the mean density of the genus. In cases lacking a value discounted 10% of the calculated value due to the absence for the genus, we used the average value for the family (Chave of leaves and branches (Delaney et al. 1998). Biomass of et al. 2006). The estimated dry biomass of each tree and palm crownless trees and dead palms was estimated by the individual was multiplied by the carbon concentrations of equation proposed by Graça et al. (1999), adopting a form 50% and 49.4%, respectively, in order to estimate carbon factor of 0.78 (Fearnside 1992). stock (Malhi et al. 2004; Goodman et al. 2013).

2.3.4 Woody debris 2.3.2 Dead biomass (necromass) For estimation dead biomass in woody debris, we adapted the For each class of dead biomass, it was necessary to determine line-intercept method described by Van Wagner (1968). Three density values for the different phases of decomposition. The 20-m lines were placed in each plot and all pieces of woody decomposition category was assigned in the field based on the debris with diameter above 10 cm that intercepted the line were classification proposed by Silva et al. (2016)(Appendix, sampled, with the exception of bamboo. The diameter of each Table 4). piece was recorded at the point it intercepted the line. A water After classification of the samples as a function of the de- content of 0.416 was considered for woody debris (Nogueira composition phase, they were separated and weighed and their et al. 2008b). To estimate the carbon stored by the woody de- individual wet volumes were determined by the water- bris, we adopted a carbon content of 46.4% (Graça et al. 1999). displacement method (Archimedes principle). Samples were in the form of cross-sectional disks in order to represent the 2.3.5 Guadua spp. biomass radial variability in wood anatomy. The dry weights of the samples were obtained after air drying for a period of 15 days Biomass of bamboo was estimated by directly weighing and the mean values for density of the samples in each of the all of the ramets found in each sampled subplot. After three decomposition classes were used for estimating the recording the number of established stems, all bamboo Annals of Forest Science (2019) 76:110 Page 5 of 13 110 ramets were sectioned using the lateral ends of the subplots as cutting limits. Weighing was done in the field using a spring balance with a maximum capacity of 200 kg and precision of 1 kg. Samples of stems (live and dead) and of leaves (live and dead) were collected for determination of water content and density. The leaves were weighed separately from the culms, and live ramets were weighed separately from dead ones. The dry weight of each weighing was calculated by deducting the water content of each sampled compartment. For conversion of bamboo biomass to carbon stock, a carbon content of 50% was adopted as a default value for vegetation in general (Brown 1997).

2.3.6 Data analysis

Fig. 4 Linear regression (R2 = 0.36, p = 0.001) between aboveground Parametric statistical tests were applied to the analysis of var- biomass (Mg ha−1) of bamboo and culm density for all sample units in iance (ANOVA) with 95% confidence for determination of the study possible effects of the impact levels on the different biomass compartments. Evaluations of the homoscedasticity and the 3 Results normality of the data were performed by the F-max of Hartley and Shapiro-Wilk tests, respectively (Hartley 1950; 3.1 Guadua spp. abundance Shapiro and Wilk 1965). We conducted all of the analyses in R software (R Core Team 2018). Two measures of bamboo abundance were evaluated: above- In cases of absence of homoscedasticity or normality in the ground biomass (Mg ha−1) and culm density per unit area data, transformation techniques (i.e., logarithmic transforma- (culms m−2). The relationship between these two variables tion) were used in order to comply with the assumptions of was observed through a significant linear regression (R2 = ANOVA. The Kruskal-Wallis non-parametric statistical test 0.36; p =0.001)(Fig.4). was applied in cases where the data did not fit the ANOVA The average dry aboveground biomass (live + dead) of assumptions (Fig. 3). Guadua spp. was 27% higher in the treatment affected by fire when compared to the control (Table 2).

Fig. 3 Schematic of the conceptual framework of the study. Cylinders represent the sampled variables. Rounded rectangles represent the methodological approach and oval frames represent the analytical steps. Double arrows show the application of a linear regression model to observe the relationship between the two abundance variables for Guadua spp. (aboveground biomass and culm density). Color figure available in online version 110 Page 6 of 13 Annals of Forest Science (2019) 76:110

Table 2 Aboveground biomass (live + dead) of trees, palms, bamboo, and forest per hectare in the different treatments

Parameter Control Fire Fire + logging p

Trees Mean (Mg ha−1)215.0a 105.4b 122.3b ANOVA 0.012* CV (%) 40.9 39.5 63.1 CI 141. 4–288.7 70.7–140.3 57.7–186.8 Palms Mean (Mg ha−1) 4.3 14.2 0.5 Kruskal-Wallis 0.197 CV (%) 84.3 253.5 116.3 CI 1.2–7.3 0–44.2 0.01–1.1 Mean (Mg ha−1)219.3a 119.7b 122.8b ANOVA Trees + palms CV (%) 41.3 36.7 63.1 CI 143.6–295.0 86.0–156,3 58.1–187.6 0.019* Mean (Mg ha−1) 38.26 48.66 41.91 ANOVA Bamboo CV (%) 44.2 33.2 52.7 CI 24.12–52.40 35.16–62.17 23.44–60.38 0.534 Mean (Mg ha−1) 259.97a 171.96b 167.34b ANOVA Forest CV (%) 36.6 24.2 51.1 CI 180.3–339.6 137.0–206.9 95.8–238.8 0.045*

Significant differences between treatments are indicated by different lower-case letters (a and b)besidethemeans;*=p <0.05 CV coefficient of variation; CI confidence interval

As was observed in the case of aboveground biomass of 3.2 Forest biomass bamboo, culm density per unit area was higher, on average, by 22% in the two treatments with fire compared to the control In addition to necromass, we sampled 1111 live trees (DBH ≥ plots (Appendix, Table 5). Although an ecologically relevant 10 cm), 105 live palms (DBH ≥ 10 cm), and 174 live and dead difference between treatments has been found here, there is a bamboos in the plots. The dry AGB (live + dead) of the forest high variability of aboveground biomass in this forest type. was obtained by summing the values for all of the biomass classes sampled in the study. There was a significant difference (p = 0.045) between the control and the disturbed treatments (Table 2). The classes that contributed most to the total amount of biomass (live + dead) were the live trees and palms, account- ing for 83.5% of the total biomass in the control treatment. Considering all treatments, live trees and palms represented 77.1% of the total dry aboveground biomass of the forest (Fig. 5). In terms of availability of combustible material, the largest amount of dead biomass was found in the treatment affected by fire, representing 17% of the total biomass in this treatment. Values for total dry aboveground biomass (live + dead) obtained in the control treatment (259.97 Mg ha−1), intermediate-impact treatment (171.96 Mg ha−1), and high- impact treatment (167.34 Mg ha−1) were converted to equivalent carbon stock values of 129.87 Mg ha−1, 85.76 Mg ha−1, and 83.57 Mg ha−1,respectively.

3.3 Biomass of live trees and palms

Fig. 5 Composition of the total (live + dead) aboveground biomass The values for live-tree biomass showed a significant dif- (Mg ha−1) of the forest for each treatment ference (p = 0.012) between the control and fire treatments. Annals of Forest Science (2019) 76:110 Page 7 of 13 110

−1 Fig. 6 Mean (+ Syx) of the aboveground biomass in Mg ha (a, b) and number (c, d) of live trees and palms per hectare by diameter class in the different treatments

This difference was detected from a Tukey test (p <0.05) differences were observed for the aboveground biomass of followed by application of ANOVA. In the case of biomass live trees + palms (p = 0.019) and for the large individuals of live palms, absence of normality in the data, even after a (p = 0.049). Using the Tukey test (p < 0.05) followed by logarithmic transformation, was shown by the Shapiro- ANOVA, differences were detected between the control Wilk test (W =0.83, p = 0.0001). This made application and impact treatments (fire and fire + logging) (Fig. 6). of the Kruskal-Wallis test necessary (Table 2). When considering the biomass of the group formed by live Considering the group formed by live trees and palms, trees and palms, we observed proximity between the values more than half of the aboveground biomass (53.7%) was obtained in the two treatments impacted by disturbances composed of large individuals (DBH > 50 cm). Significant (Table 3).

Table 3 Mean (± standard deviation) aboveground biomass Diameter class (cm) Control Fire Fire + logging ANOVA p (Mg ha−1)oflivetreesandpalms ≤ 20 13.35 ± 6.1 9.36 ± 3.32 9.97 ± 7.51 0.23** 21–30 20.05 ± 8.46 14.95 ± 3.58 15.3 ± 6.74 0.20* 31–40 25.13 ± 7.36 15.41 ± 9.76 16.45 ± 11.19 0.09* 41–50 25.36 ± 12.05 21.92 ± 18.09 20.68 ± 20.79 0.86* >50 135.43a ± 85.85 58.04b ± 38.53 60.40b ± 57.48 0.049** Total 219.3a ± 90.54 119.7b ± 44.86 122.8b ± 72.29 0.019*

Significant differences between treatments are indicated by different lower-case letters (a and b) beside the means *ANOVA & Tukey; ** Kruskal-Wallis & Mann-Whitney U-test 110 Page 8 of 13 Annals of Forest Science (2019) 76:110

4 Discussion development limited by light, remaining thinner and shorter. However, the nonlinear pattern was not conclusive for our data 4.1 Bamboo-dominated forest and disturbances set, where only one plot had high culm density with lower biomass, consistent with the nonlinear pattern (Appendix, Fig. 9). In this study, we show an increase in bamboo abundance, a Dead bamboo biomass was greater than live bamboo biomass significant decrease in large-tree aboveground biomass, and, con- for all treatments (Appendix, Fig. 7). A possible explanation for sequently, in aboveground forest biomass following fire and the large amount of dead biomass found in the current study may post-burn logging. However, one must consider the high spatial be tied to the end of the life cycle of the bamboo populations variability in aboveground biomass and different potential re- sampled. This condition was characterized in the field by the sponses of bamboo-dominated forests following anthropogenic presence of abundant fruits (Appendix, Fig. 8), a condition that disturbances. Fire occurrence in bamboo-dominated forests is frequently precedes the periodic mass mortality event of bamboo 45% higher in dead than live bamboo populations during drought populations (Janzen 1976;deCarvalhoetal.2013). years, suggesting a potential process to the maintenance and Biomass of Guadua spp. showed no significant difference expansion of bamboo in the forest (Dalagnol et al. 2018). between treatments. However, the intermediate-impact treatment Considering the last 33 years (1984–2017), the years (fire) had an average bamboo biomass more than 10 Mg ha−1 2005, 2010, and 2016 had the highest frequencies and higher than the control treatment. In another area located 27 km intensities of forest fires in the state of Acre, indicating from the city of Rio Branco and near the site of the current study, a probable relation of these phenomena to the effects of a Torezan and Silveira (2000) reported only 10.2 Mg ha−1 of bam- strong Atlantic dipole (2005), a strong Atlantic dipole boo biomass, which is approximately 4 times lower than the together with a weak El Niño (2010), and a weak mean value in the present study. In addition, the authors reported Atlantic dipole together with a strong El Niño (2016). that the biomass of Guadua spp. represented 4.2% of the total Extreme events from both of these phenomena are in- forest biomass, while in the current study the corresponding val- creasing in frequency as a consequence of anthropogenic ue is 21.5%. Considering that the bamboo ramets sampled in global warming (Meehl et al. 2007,p.779;Coxetal. both studies belonged to the same population and that the 2008;Evanetal.2009). mass-flowering event was observed in 2016, a natural increase In addition to climate change, direct human activities, such of bamboo biomass was expected due to ripening of culms and as logging, are also determinant factors for fire occurrence establishment of new ramets. (Aragão et al. 2018). At least 46% of the total forest area The mean value for aboveground biomass found by Yavit impacted by forest fires in the state of Acre is located within (2017)forGuadua spp. was 6.2 Mg ha−1 in areas located in settlement projects and at least 15% within government- the departments of Madre de Dios and Ucayali, Peru. This created conservation units, where theoretically there should value was six times lower than the mean value of be low human pressure (da Silva 2017). 38.26 Mg ha−1 for undisturbed plots in the current study. Considering all treatments, the average biomass of Guadua − 4.2 Guadua spp. abundance spp. in our study (42.96 Mg ha 1) represented 29.1% of the biomass of living trees (147.61 Mg ha−1). The largest arboreal Although the relationship between the two abundance variables individual of the current study found in the control treatment for Guadua spp. (aboveground biomass and culm density) in the was a Calophyllum brasiliensis (Guanandi) with 133.7 cm present study was positive and linear, it was not very strong DBH. The biomass estimated for this single individual was (R2 = 0.34). This relationship would be intuitive if all the ramets 26.1 Mg, which represents more than half of the estimated found in the sample units were fully sampled. However, the mean biomass of Guadua spp. per hectare. This suggests high sampling method used the lateral extremities of the subplots as uncertainty for the tree component in the absence of a much cutting limits of the culms in the biomass sampling. larger sample size, and therefore also implies high uncertainty Yavit (2017) found a nonlinear parabolic relationship between for values for bamboo when expressed as percentages of either these two variables for Guadua spp. in southeastern Peru, where, tree or total biomass. after a certain point, further increase in bamboo density did not With respect to the dominance of bamboo, the mean values − result in further increase in its biomass, instead leading to a for the control treatment (2500 culms ha 1), the intermediate- − reduction. One possible hypothesis that may explain this ecolog- impact treatment (3400 culms ha 1), and the high-impact treat- − ical mechanism is the lower density of trees in high bamboo ment (2700 culms ha 1)canallbeconsideredhighinrelationto density environments. In these scenarios, the culms lose their the values reported by Barlow et al. (2012)inAcre’sChico verticality, failing to access the forest canopy because they cannot Mendes Extractive Reserve (RESEX), where there were − rely on adjacent trees. Consequently, bamboos have their 279.7 culms ha 1 in forest with no known impact and 282.2 Annals of Forest Science (2019) 76:110 Page 9 of 13 110 culms ha−1 after forest-fire disturbance. The values in the pres- and 54% of the total aboveground biomass of the forest (live + ent study were also higher than the 1420 culms ha−1 found by dead), indicating the great importance that maintaining these Torezan and Silveira (2000) in an area near our study site, and individuals has for the forest carbon stock. the 1350 culms ha−1 reported by Yavit (2017)insoutheastern Considering only live trees and palms with DBH > 50 cm, it Peru. The bamboo densities that are closest to those found in was possible to observe a significant difference for biomass thisstudywerereportedbyGriscomandAshton(2006)onthe values between treatments (p = 0.049). The control treatment Tambopata River in Peru, with 3860 ± 265 culms ha−1 in forests had a mean value 55.4% higher than the treatment with logging dominated by Guadua weberbaueri and2375±618culmsha−1 and fire and 57.1% higher than the treatment impacted only by in forests dominated by Guadua sacocarpa. fire (Table 3). The values for biomass of live trees and palms (DBH > 4.3 Aboveground forest biomass 10 cm) sampled by Barlow et al. (2012) 3 years after a fire in Acre were higher than those obtained in the present study for − Since Amazon forest has great structural and floristic hetero- the forest without known impact (380 to 400 Mg ha 1) and for − geneity, a high variability in forest biomass is expected, even the forest affected by fire (260 to 280 Mg ha 1). There is for plots in the same forest type (Fearnside 2018). Although evidence for a structural difference in the abundance of senes- the soils under forests in southwestern Amazonia have rela- cent large trees in the Amazon along a gradient from east to tively high concentrations of phosphorus by Amazonian stan- west, with older individuals being more abundant in the east- dards (Quesada et al. 2010), the biomass values can be con- ern Amazon region (Chao et al. 2008). In addition, studies sidered low compared to values observed in other parts of the also suggest a possible difference in the mortality dynamics Brazilian Amazon (Nogueira et al. 2015). This can be ex- of these large trees along this same gradient (Chao et al. 2008; plained by high tree turnover, which follows a decreasing Phillips et al. 2008). Large trees are also most sensitive to dry gradient from west to east in the Amazon (Quesada et al. conditions from edge effects (Nascimento and Laurance 2004) 2012) and implies high recruitment and mortality rates. A and from experimental drought (Nepstad et al. 2007). reduced residence time can be expected for carbon in the for- Assuming that senescent large trees are more vulnerable to ests in the area of the present study. stress from drought (as well as disturbances from logging The legacy left by impacts associated with these forests and fire) than are smaller trees, large trees in the southwestern could be observed in our study more than 10 years after dis- Amazon would be more resistant to anthropogenic impacts turbance. The difference between the carbon stored in 1 ha of than trees of similar diameter in eastern Amazonia because the control treatment and in the treatment with the highest they are less senescent (Barlow et al. 2012). The data from impact was 46.3 Mg ha−1. This amount is equivalent to the the present study, however, reveal a high vulnerability of large emissions from the annual consumption of goods and services trees to these disturbances, given the observation of a subtle by approximately 17 Brazilians, 19 Peruvians, or 25 Bolivians reduction in the number of live trees and palms with DBH > in 2015 (Le Quéré et al. 2015). 50 cm (Fig. 6), which represented a reduction of 45.4% in the biomass of the intermediate-impact treatment and 44.0% in 4.4 Biomass of live trees and palms the biomass of the treatment with the greatest impact (Table 3). The biomass class of live trees and palms represented, on average, 74% of the total aboveground biomass (live + dead). 4.5 Bamboo-dominated forest, ecosystem services, Thus, the reduction of aboveground biomass value in the for- and potential for management est can be largely attributed to the reduction in the number of live trees and palms per hectare following disturbance (Fig. 5). The “cascade model” of ecosystem service generation pre- A large part of the high aboveground biomass found in the sented in 2010 by Haines-Young and Potschin (2010)sug- undisturbed treatment can be attributed to the concentration of gests a link between ecosystem services and human well-be- individuals in the highest diameter class (Fig. 6). ing. Despite the difficulty of defining and measuring ecosys- Although the treatment with the highest impact had more tem services, assigning a value to services provided by con- trees and palms per hectare than the treatment with interme- served forest is important for discussion and planning of pub- diate impact, the latter had a larger number of individuals with lic policies for conservation (Potschin and Haines-Young DBH > 40 cm, which was reflected by a higher biomass value 2016), such as reducing emissions from deforestation and for- of live trees and palms (Fig. 6). The biomass of live trees and est degradation (REDD+) (Fearnside 2012). palms with DBH above 40 cm represented, on average, 70% A complete cycle of generating and managing ecosystem of the biomass of live trees and palms in all diameter classes, services can be implemented by integrating public and private 110 Page 10 of 13 Annals of Forest Science (2019) 76:110 interests into planning and decision-making processes Compliance with ethical standards (Spangenberg et al. 2014). Because large trees are of central importance both for carbon stock and for the forest’s value for Conflict of interest The authors declare that they have no conflict of timber management, the impact on these trees due to the in- interest. crease of bamboo in conjunction with forest fires is both an Disclaimer The funding sponsors had no role in the design of the study, environmental and a commercial concern. in the collection, analyses do, or interpretation of data, in the writing of the manuscript, and in the decision to publish the results.

5 Conclusions Appendix

This study suggests a downward trend in the aboveground carbon stock of the forest with increasing level of impact. Reduction in the aboveground biomass of live trees and palms, especially in the diameter class over 50 cm, is one of the main consequences related to disturbances in open forests dominated by bamboo. These individuals are of central importance for the carbon stock, and, from the economic point of view, decrease in their abundance leads to a loss of value of the standing forest, reducing the already-low potential of these forests for management. The damage to Brazil’s Amazon forest initiated by the 2005 drought, documented here, adds one more piece of evidence of the need to contain global climate change and of the logic for Brazil to assume a leading role in these efforts.

Acknowledgments We thank the National Institute for Research in Amazonia (INPA: PRJ15.125), the National Institute of Science Fig. 7 Dry aboveground biomass of live and dead bamboo per hectare and Technology for the Environmental Services of Amazonia (INCT-SERVAMB), the National Council for Scientific and Technological Development (CNPq/PCI Program Proc. 304130/ 2013-3 and 301183/2015-5; CNPq: Proc. 304020/2010-9; 573810/2008-7), and the Foundation for Support of Research in Amazonas State (FAPEAM: Proc. 708565) for logistical support. We owe personal thanks to Sebastião Mota, Evandro Ferreira, Marcos Vinícos D’Oliveira, Aurora Miho Yanai, Jochen Schöngart, Reinaldo Imbrózio Barbosa, Nathan B. Gonçalves, and Bruce W. Nelson.

Statement on data availability The datasets generated and/or analyzed during the current study are available in the Zenodo repository (Ziccardi et al. 2019)athttps://doi.org/10.5281/zenodo.3530726

Funding information This study received financial support from National Institute for Research in Amazonia (INPA: PRJ15.125), the National Institute of Science and Technology for the Environmental Services of Amazonia (INCT-SERVAMB), the National Council for Scientific and Technological Development (CNPq/PCI Program Proc. 304130/2013-3 and 301183/2015-5; CNPq: Proc. 304020/2010-9; 573810/2008-7), and the Foundation for Support of Research in Amazonas State (FAPEAM: Proc. 708565). This article is a contribution of the Brazilian Research Network on Global Climate Change, FINEP/ Fig. 8 Abundance of fruits and seeds of Guadua spp. observed in the Rede CLIMA Grant Number 01.13.0353-00. field, indicating a mass-flowering event Annals of Forest Science (2019) 76:110 Page 11 of 13 110

Fig. 9 Nonlinear regression between bamboo aboveground biomass (AGB) and bamboo density for all sample units

Table 4 Characteristic of woody debris related to the phases of Phase of Description of characteristics wood decomposition decomposition

C1 Woody debris without noticeable deterioration, recently fallen and resistant to attack by microorganisms C2 Woody debris with few signs of fungal or insect attack, with initial deterioration C3 Woody debris at an advanced stage of deterioration, easily breaking or cracking if touched

Table 5 Guadua spp. density in −1 the different treatments Treatment Mean density (culms ha ) CV (%) Percentage error (%) p (ANOVA)

Control 2500 45.9 38.4 0.478 Fire 3400 50.3 42.1 Fire + logging 2700 63.2 52.9

CV coefficient of variation

References Aragão LE, Anderson LO, Fonseca MG, Rosan TM, Vedovato LB, WagnerFH,SilvaCVJ,JuniorCHL,AraíE,AguiarA, Barlow J, Berenguer E, Deeter M, Domingues LG, Gatti L, Gloor M, Malhi Y, Marengo J, Miller J, Philips OL, Barlow J Acre, Governo do Estado (2000) Programa Estadual de Zoneamento (2018) 21st century drought-related fires counteract the de- Ecológico-Econômico do Estado do Acre. Zoneamento Ecológico- cline of Amazon deforestation carbon emissions. Nat – Econômico: Recursos Naturais e Meio Ambiente Documento Commun 9(1):536 Final. Rio Branco: SECTMA 1:15–35 https://www.amazonia. Barlow J, Silveira JM, Mestre LA, Andrade RB, D'Andrea GC, Louzada cnptia.embrapa.br/publicacoes_estados/Acre/Fase%201/Vol.201_ J, Vaz-de-Melo FZ, Numata I, Lacau S, Cochrane MA (2012) Recursos_Naturais_e_Meio_Ambiente.pdf 110 Page 12 of 13 Annals of Forest Science (2019) 76:110

Wildfires in bamboo-dominated Amazonian forest: impacts on Fearnside PM (2012) Brazil's Amazon forest in mitigating global aboveground biomass and biodiversity. PLoS One 7(3):art. warming: unresolved controversies. Clim Pol 12(1):70–81. https:// e33373. https://doi.org/10.1371/annotation/8013f95e-71f4-4ec0- doi.org/10.1080/14693062.2011.581571 afe9-00cea6627685 Fearnside PM (2018) Brazil’s Amazonian forest carbon: the key to south- Brazil, IBGE (Instituto Brasileiro de Geografia e Estatística). (2017). ern Amazonia’s significance for global climate. Reg Environ Chang Diretoria de Pesquisas, Coordenação de População e Indicadores 18(1):47–61. https://doi.org/10.1007/s10113-016-1007-2 Sociais, Estimativas da população residente com data de referência Ferrante L, Fearnside PM (2019) Brazil’s new president and “ruralists” 1° de julho de 2017. Accessed 12 Feb. 2018 https://cidades.ibge. threaten Amazonia’s environment, traditional peoples and the global gov.br/xtras/perfil.php?lang=&codmun=120040&search=acre|rio- climate. Environ Conserv. https://doi.org/10.1017/ branco S0376892919000213 Brown S (1997) Estimating biomass and biomass change of tropical Ferreira EJL (2014) O bambu é um desafio para a conservação e o manejo forests: a primer. FAO Forestry Paper 134, Rome, Italy de florestas no sudoeste da Amazônia. Ciência e Cultura 66(3):46– Brown IF, Martinelli LA, Thomas WW, Moreira MZ, Ferreira CC, 51 http://cienciaecultura.bvs.br/scielo.php?pid=S0009- Victoria RA (1995) Uncertainty in the biomass of Amazonian for- 67252014000300015&script=sci_arttext ests: an example from Rondônia, Brazil. For Ecology and Goodman RC, Phillips OL, del Castillo TD, Freitas L, Cortese ST, – Management 75(1):175 189. https://doi.org/10.1016/0378- Monteagudo A, Baker TR (2013) Amazon palm biomass and allom- 1127(94)03512-U etry. For Ecol Manag 310:994–1004. https://doi.org/10.1016/j. Chao KJ, Phillips OL, Gloor E, Monteagudo A, Torres-Lezama A, foreco.2013.09.045 Martínez RV (2008) Growth and wood density predict tree mortality Graça PMLA, Fearnside PM, Cerri CC (1999) Burning of Amazonian – in Amazon forests. J Ecol 96(2):281 292. https://doi.org/10.1111/j. forest in Ariquemes, Rondônia, Brazil: biomass, charcoal formation 1365-2745.2007.01343.x and burning efficiency. For Ecol Manag 120(1–3):179–191. https:// Chave J, Muller-Landau HC, Baker TR, Easdale TA, Steege HT, Webb doi.org/10.1016/S0378-1127(98)00547-7 CO (2006) Regional and phylogenetic variation of wood density Griscom BW, Ashton PMS (2006) A self-perpetuating bamboo distur- – across 2456 Neotropical tree species. Ecol Appl 16(6):2356 2367. bance cycle in a Neotropical forest. J Trop Ecol 22(5):587–597. https://doi.org/10.1890/1051-0761(2006)016[2356:RAPVOW]2.0. https://doi.org/10.1017/S0266467406003361 CO;2 Haines-Young R, Potschin M (2010) The links between biodiversity, Cox PM, Harris PP, Huntingford C, Betts RA, Collins M, Jones CD, Jupp ecosystem services and human well-being. Ecosystem Ecology: a TE, Marengo JA, Nobre CA (2008) Increasing risk of Amazonian new synthesis 1:110–139. https://doi.org/10.1017/ drought due to decreasing aerosol pollution. Nature 45:212–215. CBO9780511750458.007 https://doi.org/10.1038/nature06960 Hartley HO (1950) The Maximum F-Ratio as a Short-Cut Test for d’Oliveira MVN, de Guarino ES, Oliveira LC, Ribas LA, Acuña MHA Heterogeneity of Variance. Biometrika 37(3/4):308 (2013) Can forest management be sustainable in a bamboo domi- Janzen DH (1976) Why bamboos wait so long to flower. Annu Rev Ecol nated forest? A 12-year study of forest dynamics in western Syst 7:347–391. https://doi.org/10.1146/annurev.es.07.110176. Amazon. Forest Ecology and Management 310:672–679. https:// 002023 doi.org/10.1016/j.foreco.2013.09.008 da Silva SS (2017) Dinâmica dos incêndios florestais no estado do Acre. Le Quéré C, Moriarty R, Andrew RM, Peters GP, Ciais P, Friedlingstein Dissertation, National Institute for Research in Amazonia https:// P, Boden TA (2015) Global carbon budget 2014. Earth System – bdtd.inpa.gov.br/handle/tede/2386 Science Data 7(1):47 85. https://doi.org/10.5194/essd-7-47-2015 da Silva SS, Fearnside PM, Graça PMLA, Brown IF, Alencar A, de Melo Lewis SL, Brando PM, Phillips OL, van der Heijden GM, Nepstad D – AWF (2018) Dynamics of forest fires in the southwestern Amazon. (2011) The 2010 Amazon drought. Science 331(6017):554 554. For Ecol Manag 424:312–322. https://doi.org/10.1016/j.foreco. https://doi.org/10.1126/science.1200807 2018.04.041 Malhi Y,Baker TR, Phillips OL, Almeida S, Alvarez E, Arroyo L, Killeen Daily GC (1997) Nature’s services. Island Press, Washington, DC TJ (2004) The above-ground coarse wood productivity of 104 – Dalagnol R, Wagner F H, Galvão L S, Nelson B W, de Aragão L E O C Neotropical forest plots. Glob Chang Biol 10(5):563 591. https:// (2018). Life cycle of bamboo in the southwestern Amazon and its doi.org/10.1111/j.1529-8817.2003.00778.x relation to fire events. Biogeosciences 15(20): 6087-6104. doi.org/ Meehl GA, Stocker TF, Collins WD, Friedlingstein P, Gaye AT, Gregory 10.5194/bg-15-6087-2018 JM, Kitoh A, Knutti R, Murphy JM, Noda A, Raper SCB, Watterson de Carvalho A L, Nelson B W, Bianchini M C, Plagnol D, Kuplich T M, IG, Weaver AJ, Zhao ZC (2007) Global climate projections. In: Daly D C. (2013). Bamboo-dominated forests of the southwest Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, Amazon: detection, spatial extent, life cycle length and flowering Tignor M, Miller HL (eds) Climate change 2007: the physical sci- waves. PloS One 8(1): art. e54852. https://doi.org/10.1371/journal. ence basis. Contribution of Working Group to the Fourth pone.0054852 Assessment Report of the Intergovernmental Panel on Climate Delaney M, Brown S, Lugo AE, Torres-Lezama A, Quintero NB (1998) Change. Cambridge University Press, Cambridge, pp 747–845 The quantity and turnover of dead wood in permanent forest plots in https://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch10.html six life zones of Venezuela. Biotropica 30(1):2–11. https://doi.org/ Molina C M C, Galiana-Martín L (2016) Fire scenarios in Spain: a terri- 10.1111/j.1744-7429.1998.tb00364.x torial approach to proactive fire management in the context of global Evan AT, Vimont DJ, Heidinger AK, Kossin JP, Bennartz R (2009) The change. Forests 7(11): art. 273. https://doi.org/10.3390/f7110273 role of aerosols in the evolution of tropical North Atlantic Ocean Nascimento HEM, Laurance WF (2004) Biomass dynamics in temperature anomalies. Science 324:778–781. https://doi.org/10. Amazonian forest fragments. Ecol Appl 14(3):S127–S138. https:// 1126/science.1167404 doi.org/10.1890/01-6003 Fearnside PM (1990) Fire in the tropical rain forest of the Amazon basin. Nascimento HE, Laurance WF (2006) Efeitos de área e de borda sobre a In: Goldammer JG (ed) Fire in the tropical biota. Springer, Berlin & estrutura florestal em fragmentos de floresta de terra-firme após 13- Heidelberg, Germany, pp 106–116. https://doi.org/10.1007/978-3- 17 anos de isolamento [ffects of area and of edges on forest structure 642-75395-4_7 in fragments of terra firme forest after 13-17 years of isolation]. Acta Fearnside PM (1992) Forest biomass in Brazilian Amazonia: comments Amazon 36(2):183–192. https://doi.org/10.1590/S0044- on the estimate by Brown and Lugo. Interciência 17(1):19–27 59672006000200008 Annals of Forest Science (2019) 76:110 Page 13 of 13 110

Nelson B W, Oliveira A C A, Batista G T, Vidalenc D, Silveira M (2001) R Core Team (2018). R: A language and environment for statistical Modeling biomass of forests in the southwest Amazon by polar computing. R Foundation for Statistical Computing, Vienna, ordination of Landsat TM. In Proceedings, Tenth Brazilian Austria. https://www.R-project.org/. Remote Sensing Symposium. Foz de Iguaçu, Paraná, São José dos Rockwell CA, Kainer KA, d’Oliveira MVN, Staudhammer CL, Baraloto Campos, SP, Brazil: INPE. pp. 1683–1690 C (2014) Logging in bamboo-dominated forests in southwestern Nepstad DC, Verssimo A, Alencar A, Nobre C, Lima E, Lefebvre P, Amazonia: caveats and opportunities for smallholder forest manage- Schlesinger P, Potter C, Moutinho P, Mendoza E, Cochrane M ment. For Ecol Manag 315:202–210. https://doi.org/10.1016/j. (1999) Large-scale impoverishment of Amazonian forests by log- foreco.2013.12.022 ging and fire. Nature 398:505–508. https://doi.org/10.1038/19066 Shapiro SS, Wilk MB (1965) An Analysis of Variance Test for Normality Nepstad DC, Tohver IM, Ray D, Moutinho P, Cardinot G (2007) (Complete Samples). Biometrika 52(3/4):591 Mortality of large trees and lianas following experimental drought Silva LFSG, de Castilho CV,de Oliveira CC, Pimentel TP, Fearnside PM, in an Amazon forest. Ecology 88(9):2259–2269. https://doi.org/10. Barbosa RI (2016) Production and stock of coarse woody debris 1890/06-1046.1 across a hydro-edaphic gradient of oligotrophic forests in the north- Nogueira EM, Fearnside PM, Nelson BW, França MB (2007) Wood ern Brazilian Amazon. For Ecol Manag 364:1–9. https://doi.org/10. density in forests of Brazil’s ‘arc of deforestation’: implications for 1016/j.foreco.2015.12.045 biomass and flux of carbon from land-use change in Amazonia. For Silveira M (2005) A floresta aberta com bambu no sudoeste da Ecol Manag 248(3):119–135. https://doi.org/10.1016/j.foreco.2007. Amazônia: Padrões e processos em múltiplas escalas. [The open 04.047 bamboo forest in southwest Amazonia: patterns and processes on Nogueira EM, Fearnside PM, Nelson BW, Barbosa RI, Keizer EWH multiple scales]. Rio Branco: Edufac. 127 pp. (2008a) Estimates of forest biomass in the Brazilian Amazon: new Smith M, Nelson BW (2011) Fire favours expansion of bamboo- allometric equations and adjustments to biomass from wood-volume dominated forests in the south-west Amazon. J Trop Ecol 27(1): – inventories. For Ecol Manag 256(11):1853 1867. https://doi.org/10. 59–64. https://doi.org/10.1017/S026646741000057X 1016/j.foreco.2008.07.022 Sombroek W (2001) Spatial and temporal patterns of Amazon rainfall: Nogueira EM, Fearnside PM, Nelson BW (2008b) Normalization of consequences for the planning of agricultural occupation and the wood density in biomass estimates of Amazon forests. For Ecol protection of primary forests. Ambio 30(7):388–396. https://doi. – Manag 256(5):990 996. https://doi.org/10.1016/j.foreco.2008.07. org/10.1579/0044-7447-30.7.388 022 Spangenberg JH, von Haaren C, Settele J (2014) The ecosystem service Nogueira EM, Yanai AM, Fonseca FO, Fearnside PM (2015) Carbon cascade: further developing the metaphor. Integrating societal pro- stock loss from deforestation through 2013 in Brazilian Amazonia. cesses to accommodate social processes and planning, and the case – Glob Chang Biol 21(3):1271 1292. https://doi.org/10.1111/gcb. of bioenergy. Ecol Econ 104:22–32 12798 Torezan JMD, Silveira M (2000) The biomass of bamboo (Guadua Oliveira A C A (2000) Efeitos do bambu Guadua weberbaueri Pilger weberbaueri Pilger) in open forest of the southwestern Amazon. sobre a fisionomia e estrutura de uma floresta no sudoeste da Ecotropica 6(1):71–76 Amazônia. Masters Thesis in Ecology, Manaus, Amazonas, Brazil: Van Wagner CE (1968) The line intersect method in forest fuel sampling. INPA/UA. 103 pp. For Sci 14(1):20–26 Phillips OL, Lewis SL, Baker TR, Chao KJ, Higuchi N (2008) The Vasconcelos SS, Fearnside PM, Graça PMLA, Nogueira EM, Oliveira changing Amazon forest. Philosophical Transactions of the Royal LC, Figueiredo EO (2013) Forest fires in southwestern Brazilian Society of London B: Biological Sciences 363(1498):1819–1827. Amazonia: Estimates of area and potential carbon emissions. https://doi.org/10.1098/rstb.2007.0033 Forest Ecology and Management 291:199–208. https://doi.org/10. Potschin M, Haines-Young R (2016) Defining and measuring ecosystem 1016/j.foreco.2012.11.044 services. In Routledge Handbook of Ecosystem Services (pp. 25-44). Veldman JW, Putz FE (2011) Grass-dominated vegetation, not species- Routledge, Taylor & Francis Group, London; New York. https://doi. diverse natural savanna, replaces degraded tropical forests on the org/10.4324/9781315775302 southern edge of the Amazon Basin. Biol Conserv 144:1419– Quesada CA, Lloyd J, Schwarz M, Patião S, Baker TR, Czimczik C, 1429. https://doi.org/10.1016/j.biocon.2011.01.011 Santos AJ (2010) Variations in chemical and physical properties of Amazon forest soils in relation to their genesis. Biogeosciences 7(5): Yavit N M (2017) Allometric derivation and estimation of Guadua 1515–1541. https://doi.org/10.5194/bg-7-1515-2010 weberbaueri and G. sarcocarpa biomass in the bamboo- Quesada CA, Phillips OL, Schwarz M, Czimczik CI, Baker TR, Patiño S, dominated forests of SW Amazonia. bioRxiv, preprint: 129262. Fyllas NM, Hodnett MG, Herrera R, Almeida S, Alvarez Dávila E, https://doi.org/10.1101/129262 Arneth A, Arroyo L, Chao KJ, Dezzeo N, Erwin T, Di Fiore A, Zanne A E, Lopez-Gonzalez G, Coomes D A, Ilic J, Jansen S, Lewis S L, Higuchi N, Honorio Coronado E, Jimenez EM, Killeen T, Lezama Miller R B, Swenson N G, Wiemann M C, Chave J (2009) Global AT, Lloyd G, López-González G, Luizão FJ, Malhi Y, Monteagudo wood density database. http://datadryad.org/resource/doi:10.5061/ A, Neill DA, Núñez Vargas P, Paiva R, Peacock J, Peñuela MC, dryad.23 Peña Cruz A, Pitman N, Priante Filho N, Prieto A, Ramírez H, Ziccardi L, Graça P M L A, Figueiredo E O, Fearnside P M (2019) Data Rudas A, Salomão R, Santos AJB, Schmerler J, Silva N, Silveira set (2016) Bamboo-dominated forest in Amazonia. V1. Zenodo. M, Vásquez R, Vieira I, Terborgh J, Lloyd J (2012) Basin-wide [Dataset]. https://doi.org/10.5281/zenodo.3530726 variations in Amazon forest structure and function are mediated by both soils and climate. Biogeosciences 9:2203–2246. https://doi. Publisher’snoteSpringer Nature remains neutral with regard to jurisdic- org/10.5194/bg-9-2203-2012 tional claims in published maps and institutional affiliations.